Nous Research strips Chinese state framing from a Qwen AI model
Nous Research has released Arete-27B, a version of Alibaba's open Qwen3.8-27B model. It is trained to drop Chinese state framing from Qwen's answers. On Nous's own test of 803 prompts, answers with propaganda fell from 22% to 0.9%. Nous says its skills are otherwise unchanged. The weights, test and training data are all public.
Where this came from. One outlet reported this. We read it, then wrote the sixty words above.